Ning Li

Conventional sea-surface density (SSD) retrieval methodologies primarily depend on statistical analysis and numerical simulations driven by in situ measurements and satellite observations. These approaches exhibit inherent limitations in capturing the patterns of data variation, utilizing feature information, and coupling with oceanographic physical mechanisms, thereby constraining prediction accuracy. To address these challenges, we propose an integrated framework that synergistically combines satellite remote sensing with deep learning through a spatiotemporal feature decomposition method and fusion strategy. First, a variational mode decomposition (VMD)-based data processing method is developed to decouple the main feature into three components that align with the real-world SSD change patterns: seasonal trends, seasonal periods, and residuals. Subsequently, we construct a feature fusion strategy guided by SSD physical laws, utilizing three-dimensional convolution to hierarchically fuse multiphase features through the VMD constraint and feature polynomial fitting. The strategy further merges trend feedback based on auxiliary features to enhance or compensate for trend information, while leveraging the attention mechanism of the transformer to strengthen key spatiotemporal information and address long-term dependencies. Comprehensive experiments demonstrate the effectiveness of the proposed model, which integrates deep learning with oceanographic principles for enhanced SSD prediction. For the Bohai Sea area, compared with the state-of-the-art spatiotemporal models, the proposed model demonstrated a significant 16.76% decrease in the root mean square error (RMSE). Compared with the conventional inversion technique, the RMSE of the proposed model was dramatically reduced by 68.20%, further validating its superiority in improving prediction accuracy.

To suppress the navigation error of the foot-mounted Pedestrian Dead Reckoning (PDR) system, many works, e.g. zero velocity-update (ZUPT) based methods, have been done without the assistance of the external sensors, but they all only can work well in the stance phase. To further improve the performance of PDR system in the swing phase, a novel PDR method with self-estimates of IMU biases is proposed in this paper. Differing from the existing methods in PDR, the bias of the accelerometer in the body frame, rather than navigation frame, is assumed to be constant. Then, a novel perspective to self-estimate the bias of the accelerometer in the body frame is provided, which is determined by the attitude matrix decomposition method. Finally, the attitude angles, the velocities, and the position results during the swing phase will be recalibrated by feedbacking the self-estimated biases. Field tests illustrate that the proposed method has better performance than that of the existing state-of-the-art algorithms.

Light Detection and Ranging (LiDAR) based place recognition plays a crucial role in autonomous driving and robotics systems. The feature information within LiDAR point clouds collected in urban environments is often sparse and scattered, which brings challenges to capture the essential feature by structured descriptors like matrices. In this paper, a novel descriptor called Occupancy Set is designed for encoding urban environments by introducing a set-based descriptor, based on which a new place recognition method is proposed. Specifically, an occupancy set is populated with ring and sector indices representing the occupied regions around selected landmarks, and the proportion of identical elements in occupancy sets can reflect the probability of them belonging to the same landmark. To expedite the retrieval of the same landmark, the occupancy set is further compressed into hash values, named as the Occupancy Set Key (OSK), using a Locality-Sensitive Hashing (LSH) algorithm, and the historically associated landmarks corresponding to the query scan can be quickly located by searching for OSKs in inverse dictionaries. The experimental comparisons between the proposed method and existing state-of-the-art algorithms are conducted on public KITTI and KITTI360 datasets, and the experimental results exhibit that the proposed method can operate in real-time and has better recall rates than existing methods, particularly in challenging scenarios, such as the intersections with varying paths. The codes of this paper will be open source for readers at https://github.com/ZhangZh3ng/Occupancy-Set-Key.

Information fusion is one of the key technologies in airborne multi-source navigation, where federated filter is widely used due to its simple structure. However, the sensor measurement of multi-source navigation system will become unreliable in some interferential environment, which leads to the state and measurement noise covariance matrix change over time and estimate difficultly. Inaccurate covariance matrice cause the loss of positioning accuracy in traditional federated filter. To address the problems, this paper puts forward an adaptive federal Kalman filtering algorithm based on Variational Bayesian and applies it to the INS/GPS/Terrain/Geomagnetic multi-source navigation system. Simulation and experiment results show that the proposed filter outperforms the traditional federated filter at improving the positioning accuracy under the condition of time-varying or even unknown state and measurement noise covariance matrix, which plays a positive role in UAV multi-source navigation.

The zero-velocity update (ZUPT) method is an effective technique, which has been widely used in low-cost foot-mounted inertial navigation systems. It can mitigate the accumulated error by using the fact that when the foot is stationary the velocity should be zero. However, it is hard to detect the zero-velocity phases (ZVPs) correctly in various motion modes. To deal with this problem, a novel adaptive zero velocity detection algorithm based on improved general likelihood ratio test (IGLRT) detector is proposed. The prior knowledge of the proposed IGLRT detector is the minima output value of inertial sensors, which is determined by an improved peak-valley detection method, while the prior knowledge of the traditional GLRT method is based on the hypothesis that, when the foot is stationary the magnitude of angular rates is zero and the magnitude of accelerometers is equal to the magnitude of gravitation. Also, an adaptive method is proposed to enhance the robustness of the IGLRT, where the thresholds of IGLRT can be adjusted adaptively according to the amplitude of accelerometers. Field tests show that the proposed algorithm performs better than conventional and state-of-the-art approaches in mixed motion modes.

This paper proposes a convex optimization problem, based on which the optimal strategy of false data injection (FDI) attacks is obtained to intrude machine-type-communications (MTC) networks from the perspective of an attacker, aiming to seek effective defensive measures based on a good understanding of attackers' behaviour. We consider a target tracking example, which is a typical application of MTC networks. Specifically, as a type of MTC devices, smart sensors each have the ability of perception, calculation and communication and all of them can form a sensor network. In this network, its sensors and transmission channels are vulnerable to FDI attacks, resulting in the degradation of system estimation performance. In order to maximize the estimation error covariance of MTC network, the attacker needs to decide which sensors and channels to intrude due to limited energy budget. The estimation error covariance of the MTC network is calculated, based on which a convex optimization problem to obtain the optimal attack strategy is proposed. Simulation results demonstrate that the optimal attack strategy maximizes the transient mean-square deviation and estimation error covariance of the MTC network.

Compared with traditional mechanical gears, magnetic gears have the advantages of no friction, low noise and low vibration. Although magnetic gear has a wide development prospect, there are few research papers on magnetic gear reducers. The calculation formula of the transmission torque is also more complicated, and there is a certain degree of error. This paper designs a magnetic gear reducer. Firstly, the transmission principle and characteristics of magnetic gear reducer are analyzed. Then the torque calculation formula is deduced and simplified. The three-dimensional model is constructed and the relevant simulation analysis is completed. Finally, the physical construction and experimental analysis are completed, and it is applied to the common life scenes of tower cranes, which proved the practical feasibility of the design.

This paper proposes a novel secure diffusion Kalman filter (dKF) algorithm to improve the estimation performance crippled by false data injection attacks on sensors in wireless sensor networks (WSNs). Different from the conventional dKF, each adjacent node in the WSNs is detected to ascertain its trustworthiness before local estimate fusion, so as to form a new secure network topology. Then the combination step is performed to fuse the information collected from the secure topology. The proposed secure dKF algorithm, having a better estimation performance, is robust to false data injection attacks on multiple sensors and partial elements of measurements. For the proposed secure dKF algorithm, its mean and mean‐square performance are derived, based on which its convergence is analysed. Additionally, the estimating and tracking problem of projectile position is investigated to confirm the effectiveness of the proposed secure dKF algorithm. It is shown by simulations that the proposed secure dKF algorithm achieves a significant estimation performance gain.

Step detection is very important for the foot-mounted Pedestrian Dead Reckoning (PDR) system. To deal with the false detection caused by variant stepping modes and occasional disturbance, a novel robust step detection algorithm is proposed. The proposed algorithm detects the number of steps and zero-velocity phase using data from inertial sensors mounted on the user's foot. It includes two parts: a novel anti-fault detection algorithm and a novel zero-velocity detection method, which are used to reduce the over-detection and miss-detection of steps, respectively. Field tests show that the proposed approach is robust under varying motion speeds in different step modes for different persons and the average accuracy of the proposed algorithm is 99.94% in all motion modes, which significantly exceeds conventional estimation approaches in terms of step detection accuracy.